To determine in patients with a cervical length < 25 mm on transvaginal ultrasound if the severity of cervical length shortening can be used to predict preterm premature rupture of membranes (PPROM) and the contribution of PPROM to preterm delivery in these patients. We retrospectively reviewed asymptomatic singleton pregnancies between 14 and 24 weeks at high risk of preterm delivery by obstetric history and transvaginal cervical length < 25 mm. Cases developing subsequent PPROM were compared with controls for transvaginal sonographic cervical characteristics. Of 69 patients identified to have a cervical length < 25 mm, 27 (39%) had PPROM, and 42 (61%) did not. Mean +/- standard deviation (SD) cervical length was 12.7 +/- 8.7 mm and 17.0 +/- 7.6 mm in the two groups, respectively (P = 0.04). Mean +/- SD cervical funneling was 57.4 +/- 31.4% and 40.0 +/- 28.1%, respectively (P = 0.01). The characteristics most predictive of PPROM were: cervical length of < 10 mm (sensitivity, specificity, positive and negative predictive values of 33, 90, 69, and 68%, respectively; odds ratio, 4.8; 95% confidence interval, 1.3-17.5) and cervical funneling > 75% (sensitivity, specificity, positive and negative predictive values of 33, 93, 75 and 68%, respectively; odds ratio, 6.5; 95% confidence interval, 1.6-26.9). Stepwise logistic regression revealed cervical length to be a significant predictor of PPROM (odds ratio, 4.0; 95% confidence interval, 1.1-14.2). In patients at high risk for preterm delivery because of obstetric history and transvaginal sonographic cervical length < 25 mm, a cervical length < 10 mm and cervical funneling > 75% were most predictive of PPROM. PPROM was the major contributor to preterm delivery in these patients.